Manual failover from standby node in ClusterXL

[Tomas_Hamrle]

Hello,

I had a situation with two 1450 in ClusterXL. Once the active member started to drop management traffic - SSH, HTTPS and connection from management server. Ping worked and all traffic passed correctly from LAN to WAN. The only issue was, that I could not connect to gateway. The reboot resolved the issue. I wanted to switch the traffic to standby gateway, but I could not connect to active gateway and perform the failover with command clusterXL_admin.csh down.

Is it possible to manually force the stadby member of ClusterXL to become active member (GAiA Embedded or GAiA)?

Thank you

[G_W_Albrecht]

Yes, this is documented very well: sk55081: Best Practices - Manual fail-over in ClusterXL

For GAiA Embedded we have sk65060: How to cause a manual fail-over in ClusterXL on SG-8x appliances.

[Tomas_Hamrle]

Well, maybe I'm blind, but I did not find solution for situation such as:

GW1 - Active, GW2 - Standby

I cannot connect to GW1. I can connect to GW2 and I want GW2 to become active.

[G_W_Albrecht]

Afaik issuing # clusterXL_admin up -p on GW2 will make it active...

[Tomas_Hamrle]

Well it does not work as I expect. Standby GW is still standby:

GW2# cphaprob state

Cluster Mode: High Availability (Active Up) with IGMP Membership

ID Unique Address Assigned Load State Name

1 172.16.0.1 100% ACTIVE chkp-1
2 (local) 172.16.0.2 0% STANDBY chkp-2

GW2# clusterXL_admin up -p
Setting member to normal operation ...
Member current state is STANDBY

GW2# clusterXL_admin down
Setting member to administratively down state ...
Member current state is DOWN

GW2# clusterXL_admin up -p
Setting member to normal operation ...
Member current state is STANDBY

GW2# cphaprob state

Cluster Mode: High Availability (Active Up) with IGMP Membership

ID Unique Address Assigned Load State Name

1 172.16.0.1 100% ACTIVE chkp-1
2 (local) 172.16.0.2 0% STANDBY chkp-2

[G_W_Albrecht]

That was a nice guess but it seems in this case you have to take down GW1 manually using power off - or contact TAC for a better solution...

[Maarten_Sjouw]

Just another reason to like VRRP.... 😉